Motion vector detection device

1. A motion vector detection device comprising: hardware, including a processor; an edge extraction unit implemented at least by the hardware and that extracts edge points from each of a current frame and a previous frame, the previous frame being a frame preceding the current frame; a feature point extraction unit implemented at least by the hardware and that extracts, from each of the previous frame and the current frame, a feature point which is an edge point whose positional relationship with another at least one edge point within a frame is uniquely determined within the frame, among a plurality of edge points extracted by the edge extraction unit; a feature point pair extraction unit implemented at least by the hardware and that extracts, as a feature point pair, a pair of feature points, the positional relationship of which is common, among pairs of a feature point of the previous frame and a feature point of the current frame; a motion vector calculation unit implemented at least by the hardware and that calculates, as a motion vector, a difference between coordinate values of the feature point pair extracted by the feature point pair extraction unit, wherein the feature point extraction unit extracts the feature point using a hash value of a distance between edge points as the positional relationship, including a hash table for each frame, the hash table having registered therein a group of a hash value, an appearance frequency, and a final appearance coordinate value, without a list of appearance coordinate values, of an edge point having the hash value, wherein the edge extraction unit calculates the hash value for each edge point extracted from the previous frame and the current frame, updates the appearance frequency corresponding to the hash value in the hash table, and stores the appearance coordinate of the edge point to the final appearance coordinate value corresponding to the hash value, the feature point extraction unit extracts, as a feature point, an edge point paired with a hash value corresponding to an appearance frequency of 1 in each of the hash table of the current frame and the hash table of the previous frame, and the feature point pair extraction unit extracts, from the previous frame and the current frame, a feature point pair that satisfies a condition that all absolute values of differences between pixel values of each component of the feature points pairs are equal to or smaller than a threshold, in addition to a condition of having the same positional relationship with another edge point.

2. The motion vector detection device according to claim 1 , wherein the edge extraction unit extracts, as an edge point, a point where each difference between pixel values of adjacent pixels in two directions perpendicular to each other exceeds a threshold.

3. The motion vector detection device according to claim 1 , wherein the feature point extraction unit extracts the feature point using a distance between edge points as the positional relationship.

4. The motion vector detection device according to claim 1 , including a hash table for each frame, the hash table having registered therein a group of a hash value, an appearance frequency, a final appearance coordinate value, and a pixel value of an edge point having the hash value, wherein the edge extraction unit calculates the hash value for each edge point extracted from the previous frame and the current frame, updates the appearance frequency corresponding to the hash value in the hash table, stores the appearance coordinate of the edge point to the final appearance coordinate value corresponding to the hash value, and stores the pixel value of the edge point to a pixel value corresponding to the hash value, the feature point extraction unit extracts, as a feature point, an edge point paired with a hash value corresponding to an appearance frequency of 1 in each of the hash table of the current frame and the hash table of the previous frame, and the extraction feature pair extraction unit extracts, as a feature point pair, a pair of feature points which have the same positional relationship and at which an absolute value of a difference between pixel values is equal to or smaller than a threshold, among pairs of a feature point of the previous frame and a feature point of the current frame.

5. The motion vector detection device according to claim 1 , wherein the feature point extraction unit uses one or more preceding edge points as another at least one edge point.

6. The motion vector detection device according to claim 1 , wherein the motion vector calculation unit comprises: motion vector candidate generation unit that generates, as a motion vector candidate, a difference between coordinate values for each feature point pair when a plurality of feature point pairs extracted from the previous frame and the current frame are present; and motion vector selection unit that selects, as a motion vector, a motion vector candidate having a highest appearance frequency from among motion vector candidates.

7. A motion vector detection method comprising: extracting edge points from each of a current frame and a previous frame which is a frame preceding the current frame; extracting, from each of the previous frame and the current frame, a feature point which is an edge point whose positional relationship with another at least one edge point within a frame is uniquely determined within the frame, among a plurality of edge points; extracting, as a feature point pair, a pair of feature points, the positional relationship of which is common, among pairs of a feature point of the previous frame and a feature point of the current frame; calculating, as a motion vector, a difference between coordinate values of the feature point pair, wherein the feature point is extracted using a hash value of a distance between edge points as the positional relationship, the edge is extracted by calculating, with respect to a hash table for each frame having registered therein a group of a hash value, an appearance frequency, and a final appearance coordinate value, without a list of appearance coordinate values, of an edge point having the hash value, the hash value for each edge point extracted from the previous frame and the current frame, updating the appearance frequency corresponding to the hash value in the hash table, and storing the appearance coordinate of the edge point to the final appearance coordinate value corresponding to the hash value, the feature point is extracted by extracting, as a feature point, an edge point paired with a hash value corresponding to an appearance frequency of 1 in each of the hash table of the current frame and the hash table of the previous frame, and from the previous frame and the current frame, a feature point pair is extracted that satisfies a condition that all absolute values of differences between pixel values of each component of the feature points pairs are equal to or smaller than a threshold, in addition to a condition of having the same positional relationship with another edge point.

8. The motion vector detection method according to claim 7 , wherein the edge is extracted by extracting, as an edge point, a point where each difference between pixel values of adjacent pixels in two directions perpendicular to each other exceeds a threshold.

9. The motion vector detection method according to claim 7 , wherein the feature point is extracted using a distance between edge points as the positional relationship.

10. The motion vector detection method according to claim 7 , wherein the feature point pair is extracted by extracting, from the previous frame and the current frame, a feature point pair that satisfies a condition that an absolute value of a difference between pixel values of feature points is equal to or smaller than a threshold, in addition to a condition of having the same positional relationship with another edge point.

11. The motion vector detection method according to claim 7 , wherein the edge is extracted by calculating, with respect to a hash table for each frame having registered therein a group of a hash value, an appearance frequency, a final appearance coordinate value, and a pixel value of an edge point having the hash value, the hash value for each edge point extracted from the previous frame and the current frame, updating the appearance frequency corresponding to the hash value in the hash table, storing the appearance coordinate of the edge point to the final appearance coordinate value corresponding to the hash value, and storing the pixel value of the edge point to a pixel value corresponding to the hash value, the feature point is extracted by extracting, as a feature point, an edge point paired with a hash value corresponding to an appearance frequency of 1 in each of the hash table of the current frame and the hash table of the previous frame, and the feature point pair is extracted by extracting, as a feature point pair, a pair of feature points which have the same positional relationship and at which an absolute value of a difference between pixel values is equal to or smaller than a threshold, among pairs of a feature point of the previous frame and a feature point of the current frame.

12. The motion vector detection method according to claim 7 , wherein the feature point is extracted using one or more preceding edge points as another at least one edge point.

13. The motion vector detection method according to claim 7 , wherein the motion vector is calculated by generating, as a motion vector candidate, a difference between coordinate values for each feature point pair when a plurality of feature point pairs extracted from the previous frame and the current frame are present, and selecting, as a motion vector, a motion vector candidate having a highest appearance frequency from among motion vector candidates.

14. A non-transitory computer readable medium recording a motion vector detection program for causing a computer to execute: extracting edge points from each of a current frame and a previous frame which is a frame preceding the current frame; extracting, from each of the previous frame and the current frame, a feature point which is an edge point whose positional relationship with another at least one edge point within a frame is uniquely determined within the frame, among a plurality of edge points; extracting, as a feature point pair, a pair of feature points, the positional relationship of which is common, among pairs of a feature point of the previous frame and a feature point of the current frame; calculating, as a motion vector, a difference between coordinate values of the feature point pair, wherein the feature point is extracted by extracting the feature point using a hash value of a distance between edge points as the positional relationship, a hash table for each frame, the hash table having registered therein a group of a hash value, an appearance frequency, and a final appearance coordinate value, without a list of appearance coordinate values, of an edge point having the hash value, wherein the edge is extracted by calculating the hash value for each edge point extracted from the previous frame and the current frame, updating the appearance frequency corresponding to the hash value in the hash table, and storing the appearance coordinate of the edge point to the final appearance coordinate value corresponding to the hash value, the feature point is extracted by extracting, as a feature point, an edge point paired with a hash value corresponding to an appearance frequency of 1 in each of the hash table of the current frame and the hash table of the previous frame, and from the previous frame and the current frame, a feature point pair is extracted that satisfies a condition that all absolute values of differences between pixel values of each component of the feature points pairs are equal to or smaller than a threshold, in addition to a condition of having the same positional relationship with another edge point.

15. The non-transitory computer readable medium according to claim 14 , wherein the edge is extracted by extracting, as an edge point, a point where each difference between pixel values of adjacent pixels in two directions perpendicular to each other exceeds a threshold.

16. The non-transitory computer readable medium according to claim 14 , wherein the feature point is extracted by extracting the feature point using a distance between edge points as the positional relationship.

17. The non-transitory computer readable medium according to claim 14 , wherein the feature point pair is extracted by extracting, from the previous frame and the current frame, a feature point pair that satisfies a condition that an absolute value of a difference between pixel values of feature points is equal to or smaller than a threshold, in addition to a condition of having the same positional relationship with another edge point.

18. The non-transitory computer readable medium according to claim 7 , including a hash table for each frame, the hash table having registered therein a group of a hash value, an appearance frequency, a final appearance coordinate value, and a pixel value of an edge point having the hash value, wherein the edge is extracted by calculating the hash value for each edge point extracted from the previous frame and the current frame, updating the appearance frequency corresponding to the hash value in the hash table, storing the appearance coordinate of the edge point to the final appearance coordinate value corresponding to the hash value, and storing the pixel value of the edge point to a pixel value corresponding to the hash value, the feature point is extracted by extracting, as a feature point, an edge point paired with a hash value corresponding to an appearance frequency of 1 in each of the hash table of the current frame and the hash table of the previous frame, and the feature point pair is extracted by extracting, as a feature point pair, a pair of feature points which have the same positional relationship and at which an absolute value of a difference between pixel values is equal to or smaller than a threshold, among pairs of a feature point of the previous frame and a feature point of the current frame.

19. The non-transitory computer readable medium according to claim 14 , wherein the feature point is extracted by using one or more preceding edge points as another at least one edge point.

20. The non-transitory computer readable medium according to claim 14 , wherein the motion vector is calculated by generating, as a motion vector candidate, a difference between coordinate values for each feature point pair when a plurality of feature point pairs extracted from the previous frame and the current frame are present; and selecting, as a motion vector, a motion vector candidate having a highest appearance frequency from among motion vector candidates.

21. A video signal coding device comprising: hardware, including a processor; a motion vector detection unit implemented at least by the hardware and that detects a motion vector by comparing a current frame with a previous frame, the previous frame being a frame preceding the current frame; a moving region detection unit implemented at least by the hardware and that detects, as a moving region, an identical or similar image region which exists in both the previous frame and the current frame and whose position on a screen is changed by the motion vector detected by the motion vector detection unit; a motion compensation unit implemented at least by the hardware and that copies the moving region detected by the moving region detection unit, to a destination indicated by the motion vector on the previous frame; an update region detection unit implemented at least by the hardware and that detects, as an update region, a region where the previous frame and the current frame which are obtained after motion compensation differ from each other; and a region coding unit implemented at least by the hardware and that codes as an image, the update region detected by the update region detection unit, wherein the motion vector detection device as set forth in claim 1 is used as the motion vector detection unit.

22. The video signal coding device according to claim 21 , wherein the update region detection unit comprises update region detection area setting unit that sets, on a frame, an update region detection area for detecting an update region.